Predicting Invasive Ductal Carcinoma by Using Deep Convolutional Neural Network
Shuaipeng Dong · Atlantis Highlights in Computer Sciences/Atlantis highlights in computer sciences · 2023
Due to the nature of Breast Cancer, it is challenging to make correct diagnosis based on histopathology images.And it is crucial to make early diagnosis for a complete cure.In this paper, a Neural Network algorithm was proposed to train on sets of breast histopathology images.Based on Convolutional Neural Network (CNN), it can be realized to detect and extract spatial features of images.A deep Convolutional Neural Network architecture similar to VGGNet is proposed for this study, which contains 6 3×3 layers of depth-wise Convolutional layers, 3 pooling layers and 1 fully connected layer.The proposed model was trained using Kaggle dataset of breast histopathology images, 50 epochs, with batch size of 250.The model utilizes Adagrad optimizer with learning rate of 1×10-2, decay equal to value (i.e.learning rate/number of epochs), and Binary Crossentropy as loss function.The proposed model results in 91.28% accuracy and 0.22 loss.